Mathematical Models for Anatomical Landmarks and Point Distribution Representation of Anatomical Objects for Medical Image Understanding

Abstract

One of the important purposes of the new discipline “Computational Anatomy” is robust computational understanding of clinical images. For the purpose, various models including shape models for anatomical objects are intensively and extensively studied so far. In this short paper, we introduce three related topics from our research results, which are (1) uncertainty quantification of anatomical landmarks, (2) non-rigid ICP (iterative closest point) with statistical shape model and outlier consideration, and (3) variational methods for computation of shape average for a 2D point distribution model and a grey-valued image-based shape representation. In each topic, preliminary results are shown by using clinical data sets to prove the clinical feasibility of the proposed models.